Características de la población investigada

Participants reported employing fewer strategies in the 1-back condition (mean frequency rating = 0.69) compared with the 4-back condition (mean frequency rating = 1.12). With the exception of passive listening, all other strategies were most frequently employed in 4-back than 1-back (Figure 5.5).

Non-parametric Wilcoxon signed ranks tests showed significantly greater use of strategies such as keeping track of the last four items, elaboration, and chunking, in 4-back

and the greater use of passive listening in 1-back, ps < .05. In contrast, rehearsal,

Figure 5.5 Mean frequency ratings of self-reported strategy use for each n-back condition in Experiment 4. Participants rated their use of each strategy from 0 (= almost never) to 3 (= almost always), see text for full statements.

5.5 Discussion

The objective of Experiment 4 was to identify the resource demands of serial updating in n- back and contrast these with running span demands. Using the divided attention approach, two n-back tasks were administered concurrently with the CRT task and the concurrent CRTs supplied an index of resource demands. To distinguish the demands associated with n-back updating and recognition judgements, the concurrent CRTs associated with 4-back and 1-back were compared. In line with predictions, the CRTs were considerably slower in the former. This pattern indicated greater levels of cognitive demands when participants had to maintain and update four items rather than a single one, consistent with the EEG data reported by Watter et al. (Watter et al., 2001). There was also a marked disparity in memory performances across the two tasks. Performance was significantly poorer in 4-back than 1- back in terms of both recognition accuracy and false alarms, in line with previous studies demonstrating a decrease in performance with an increase in the value of n (Jaeggi et al., 2010; Juvina & Taatgen, 2007; Kane et al., 2007; Mackworth, 1959; Oberauer, 2005; Szmalec et al., 2011).

Across both n-back tasks, the CRTs associated with the first update item were greater than the final non-update item, but the magnitude of this effect of updating did not vary between tasks. This could suggest that the activation of an updating mode when the first to-

be-updated item is encountered demands similar resources irrespective of the n-back

condition. It may also be that there is indeed a difference between the resources required to trigger an updating mode in 4-back versus 1-back, but it was not apparent in the data. This is because the CRT increase in 4-back was observed at position four, rather than position five (= n+1) at which they were anticipated.

This early elevation in processing demands was also found in Experiment 3 that examined the (self-paced) presentation times in running span. In both experiments, the variable of interest (presentation times in Experiment 3 and concurrent CRTs in Experiment 4) successively increased until position four (= n) rather than position five (= n+1). Two possible explanations considered in the context of the previous experiment could also be applied to the present data. One possibility is that participants preemptively update their target set at position four. In other words, they might discard the first encoded item from a sequence, e.g. ABCD) and continue holding the last three items in the format BCDx to allow for the fifth item to be encoded efficiently. While this may be a useful strategy in running span in which the likelihood of a recall test following position four was quite low, it is not

back, due to the recognition judgement required at position five to compare it to the first item. The second, more likely possibility is that the increase in CRTs at position four reflects a preparatory process in which participants anticipate the first recognition judgement at position five and attempt to ready their WM for a quick decision process. For instance, it may be that segmenting the sequence ABCD in the format A-BCD allows the relevant n-back to be rapidly identified and easily compared with the most recent item.

In addition to the rise in CRTs at position four, the CRTs were also elevated 600 ms from the onset of update items in both 4-back and 1-back. This was the interval in which

item presentation was still ongoing. Following this, the CRTs in 4-back, but not 1-back,

continued to increase, with a maximum level observed at a latency of approximately 1000 ms. The CRTs subsequently decreased over the remaining interval returning to the updating baseline before the onset of the next item in both 1-back and 4-back. The increase at 600 ms could reflect the recognition judgement as it was common to both n-back tasks. The specific elevation at 1000 ms specifically found in 4-back could reflect the serial updating process,

as it was more demanding in 4-back than 1-back. Together, these data suggest that the

recognition judgement in n-back occurs while stimulus is being presented, prior to memory updating.

The CRT increase at around 1000 ms in the 4-back task may also reflect resources required to resolve the conflict between a familiarity and recollection signal (Oberauer, 2005; Szmalec et al., 2011). It has been previously argued that 1-back task does not require conflict resolution as the most familiar item is likely to be the last item encountered and thereby enabled rapid access to the target 1-back item (Oberauer, 2005). In fact, it may be that in 1-back every new item simply replaces the last, and only, item held in WM without recruiting any update process. As such, it would resemble a relatively quick and effortless memory substitution or reset process, as in modified span in Experiment 1 (see also, Kessler & Meiran, 2008). This interpretation was corroborated by the strategy data, whereby active demanding strategies were more common in 4-back and passive listening in 1-back.

A comparison of the CRT functions between running span and n-back revealed no difference. This outcome favours the overlap account that the two tasks rely on the same process of updating (Bäckman et al., 2017; Chatham et al., 2011; Dahlin et al., 2008a; Ruiz & Elosúa, 2013b). However, these data are inconsistent with the finding that n-back involves an item recognition process evident 600 ms from stimulus onset. It is surprising that the recognition judgement (required only in n-back and not running span) was not apparent when CRT data were compared between the two tasks. It could be that the two components were

indistinct, perhaps even overlapping in time within 4-back, such that the statistical analyses could only detect them in comparison with 1-back but not running span.

These data warrant further investigation and highlight the need for a different means to separate the n-back task components. Instead of comparing two n-back conditions, it may also be more useful to decompose n-back into its constituent components (item comparison and memory updating) to develop a greater understanding of the task demands. Such an approach was used by Rac-Lubashevsky and Kessler (2016) who developed a novel reference-back task. As argued in the introduction of this chapter, the reference-back task however fails to capture the demands of serial updating as it only requires a single item replacement rather than systematic changes to the serial positions of already-encoded items (p. 131). A modification of the task could present a way forward. Instead of a single item target set, the reference-back could involve a target set of n items. In this case, each cycle of updating would involve the discarding of the first target item and repositioning of the remaining ones, as in typical serial updating paradigms. Future investigations could compare the demands of updating trials with those in which WM is unchanged and thus isolate the demand characteristics specific to n-back updating.

In summary, Experiment 4 found a close resemblance in the resource demands of n- back and running span, supporting the hypothesis that the two tasks rely on the same process of serial updating. The data additionally suggested that the memory recognition judgement in n-back precedes the memory update process. Suggestions for further investigations were made to clarify the timing and unique demands of the two n-back components, while also allowing more specific tests of the updating process in both n-back and running span